Upload Knowledge Base Document

Changes state - verify before retry Adds a real PDF or text FILE to a knowledge base and indexes it — the same import the dashboard upload performs. This is NOT a read-only action. Use it after create_upload_url (kind kb_document, target_id = kb_id) once the user has uploaded the file: pass the upload_id. In hosts that attach chat files to tools (ChatGPT), the host fills in file for a document the user attached; leave it alone otherwise. For text you already have in the conversation use upload_knowledge_base_text instead — never base64-encode a PDF yourself. Titles must be unique within the knowledge base.

Parameters

Name	Type	Required	Description
`kb_id`	`string`	Yes	The knowledge base ID, e.g. kb_abc123
`title`	`string`	Yes	Document title, unique within the knowledge base
`description`	`string`	No	Optional one-line description of the document
`upload_id`	`string`	No	Only for a file on a disk YOU control that is too large for file_text (over ~50 KB): call create_upload_url with reveal_url: true, PUT the bytes yourself from a machine with open internet access, then pass the upload_id here. Never for a file the user attached in the chat (pass it as `file`), and never from a hosted chat sandbox such as ChatGPT code interpreter, which cannot reach the storage host.
`file`	`object`	No	ALWAYS use this when the user attached a file in the chat: pass the attached file here and the host turns it into a download link for the server. Fastest path, any size. NEVER open, convert or base64-encode an attached file yourself.
`file_name`	`string`	No	File name with extension (.pdf or .txt). Only needed with file_base64; taken from the upload or attachment otherwise.
`file_base64`	`string`	No	Base64 content for a SMALL .pdf/.txt only. Prefer upload_id or the host attachment for anything real.

Complete input schema

{
  "type": "object",
  "properties": {
    "kb_id": {
      "type": "string",
      "description": "The knowledge base ID, e.g. kb_abc123"
    },
    "title": {
      "type": "string",
      "minLength": 1,
      "maxLength": 150,
      "description": "Document title, unique within the knowledge base"
    },
    "description": {
      "type": "string",
      "maxLength": 500,
      "description": "Optional one-line description of the document"
    },
    "upload_id": {
      "type": "string",
      "description": "Only for a file on a disk YOU control that is too large for file_text (over ~50 KB): call create_upload_url with reveal_url: true, PUT the bytes yourself from a machine with open internet access, then pass the upload_id here. Never for a file the user attached in the chat (pass it as `file`), and never from a hosted chat sandbox such as ChatGPT code interpreter, which cannot reach the storage host."
    },
    "file": {
      "type": "object",
      "properties": {
        "file_id": {
          "type": "string",
          "description": "Host-side file id"
        },
        "download_url": {
          "type": "string",
          "description": "Short-lived signed URL the host provides for the attached file"
        },
        "file_name": {
          "type": "string"
        },
        "mime_type": {
          "type": "string"
        },
        "size": {
          "type": "number"
        }
      },
      "required": [
        "file_id",
        "download_url"
      ],
      "additionalProperties": false,
      "description": "ALWAYS use this when the user attached a file in the chat: pass the attached file here and the host turns it into a download link for the server. Fastest path, any size. NEVER open, convert or base64-encode an attached file yourself."
    },
    "file_name": {
      "type": "string",
      "description": "File name with extension (.pdf or .txt). Only needed with file_base64; taken from the upload or attachment otherwise."
    },
    "file_base64": {
      "type": "string",
      "description": "Base64 content for a SMALL .pdf/.txt only. Prefer upload_id or the host attachment for anything real."
    }
  },
  "required": [
    "kb_id",
    "title"
  ],
  "additionalProperties": false,
  "$schema": "http://json-schema.org/draft-07/schema#"
}

Registered output schema

{
  "type": "object",
  "properties": {
    "document_id": {
      "description": "ID of the created document"
    },
    "knowledge_base_id": {
      "type": "string",
      "description": "The knowledge base it was added to, prefixed kb_"
    },
    "title": {
      "type": "string",
      "description": "The document title as stored"
    },
    "file_name": {
      "type": "string",
      "description": "The file name that was imported"
    },
    "bytes": {
      "type": "number",
      "description": "Size of the imported file"
    },
    "source": {
      "type": "string",
      "description": "How the file arrived: 'upload', 'attachment' or 'base64'"
    },
    "status": {
      "type": "string",
      "description": "'processing' while the file is being indexed, 'ready' once searchable, 'failed' if indexing did not complete"
    },
    "note": {
      "type": "string",
      "description": "What happens next"
    }
  },
  "required": [
    "knowledge_base_id",
    "title",
    "file_name",
    "bytes",
    "source",
    "status",
    "note"
  ],
  "additionalProperties": true,
  "$schema": "http://json-schema.org/draft-07/schema#"
}

Before you call

Connect and initialize the MCP client with OAuth or an API key for the intended workspace.
Provide the required parameters: kb_id, title.
If your OAuth grant covers several workspaces, also pass organization_id with the workspace to act in, from List My Organizations.
Provide one file source: a host attachment, a completed upload_id, or supported inline content. The optional file fields are alternatives; omitting all of them does not submit a file. See MCP file uploads.

Example MCP request

Send tools/call after your client completes MCP initialization. Replace placeholder values with IDs and inputs from your workspace. For object and array parameters, follow the nested requirements in the parameter description instead of treating an empty value as complete.

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "upload_knowledge_base_document",
    "arguments": {
      "kb_id": "kb_abc123",
      "title": "Example resource",
      "upload_id": "COMPLETED_UPLOAD_ID"
    }
  }
}

Successful results include structuredContent and a text content block containing the same JSON. Check isError first, then read structuredContent or parse content[0].text. See MCP responses and errors for the full envelope and failure behavior.

Verify the result

Confirm the response is not marked isError and read the document’s status from the response. No MCP tool reads a document’s status afterwards: the knowledge base tools return knowledge bases, not their documents. Check processing in the dashboard or with the REST List Knowledge Base Documents endpoint.

Retry safety

This tool changes workspace state. After a timeout, read the affected resource before retrying so you do not create duplicate or conflicting changes.

Upload Knowledge Base Text: upload_knowledge_base_text
Create Upload URL: create_upload_url
List Knowledge Base Documents: list_knowledge_base_documents
Get Knowledge Base Document: get_knowledge_base_document

Start Here

Tool Reference

Parameters

Complete input schema

Registered output schema

Before you call

Example MCP request

Verify the result

Retry safety

​Parameters

​Complete input schema

​Registered output schema

​Before you call

​Example MCP request

​Verify the result

​Retry safety

​Related MCP tools

​Related guides

Parameters

Complete input schema

Registered output schema

Before you call

Example MCP request

Verify the result

Retry safety

Related MCP tools

Related guides